Delegating work to AI is associated with more after-hours rumination among hybrid and remote workers
A two-wave study of hybrid and predominantly remote knowledge workers who use AI at work at least weekly found that greater reliance on AI delegation was associated with more work-related rumination outside working hours and poorer recovery. The mechanism the authors identify is “AI process vigilance”, meaning the mental effort of monitoring what the tool is doing, which predicted rumination strongly (β = 0.411 on that path).
The study, by Zhang, Li and Chen, was published in Frontiers in Psychology on 24 September 2026, with a seven-week interval between waves and 642 retained respondents, a retention rate of 86.5%.
Why this matters
The assumption behind handing a task to an AI tool is that the task leaves your head. This finding says that for a meaningful share of people it does the opposite: the work goes, the supervision stays, and the supervision is what follows you into the evening. Checking whether the output was right, whether the tool misunderstood the brief, whether something plausible-looking was wrong, all of that is cognitive load that does not appear on a timesheet.
It lands hardest on remote and hybrid workers because the boundary that ends the supervising is a physical one for office workers and a decided one for everybody else. If you work from home and use AI tools heavily, the recovery problem is worth treating as a real one rather than a discipline failure, and the practical lever is deciding in advance which outputs you will check and which you will not.
Context
This is a self-report survey design, so the findings are associations rather than causes, and they should be described that way. The two-wave structure and the high retention make it stronger than a single-point survey, but it cannot establish that AI delegation causes poorer recovery.
It sits alongside the growing evidence that remote work’s health effects run in both directions at once, and alongside the EU-level push to put psychosocial risk assessment on a statutory footing.
What to watch
Whether the finding replicates in a European sample with objective recovery measures rather than self-report.